Automatic Detection of Synthetic Biomedical Images Using Convolutional Block Attention Module-Based Deep Neural Network

With the advancements in deep learning technologies in recent years, the generation of synthetic images, videos, and sounds has become easier. In the medical field, this can manipulate biomedical images, which may lead to incorrect diagnoses by doctors. These disturbances in the medical domain can lead to financial and life-threatening problems for hospitals. Deepfake detection is crucial to safeguard patients' health by ensuring the trustworthiness of medical diagnoses and to improve healthcare outcomes in an era dominated by digital technologies. This study aims to develop an effective deep learning method by integrating Convolutional Neural Networks (CNN) with Convolution Block Attention Module (CBAM) to detect synthetic medical images. The CBAM architecture helps to enhance the model's ability to focus on relevant features and improve its performance. The dataset used comprises real and synthetic images of Lung CT scans. In the comparative study performed with other state-of-the-art models, the proposed model outperformed other methods with an accuracy of 97.12% and a recall value of 99.05%. Hence, this study shows the effectiveness of attention mechanisms not only in improving deepfake detection accuracy but also in terms of computational efficiency, making them useful for real-world applications. The findings show that the model proposed in this work can be used for fast and accurate medical deepfake detection.